CRD summary

This review compared the accuracy of commercial assays for detecting Clostridium difficile toxin in stools. The conclusion that where prevalence of CDT was low (<10%), the positive predictive value of the assays was unacceptably low, is a reasonable interpretation of the data, but should be viewed with caution given the methodological limitations of the review.

Authors' objectives

To compare the accuracy of commercial immunoassays for the detection of Clostridium difficile toxin (CDT) A and B in stool samples.

Searching

MEDLINE and EMBASE were searched; search terms were reported and included terms for condition, test and diagnostic accuracy. Databases from the American Society for Microbiology, European Congress of Clinical Microbiology and Infectious Diseases conferences were searched for relevant abstracts. Bibliographies of included studies were screened for additional articles. Searches covered the period from 1994 to November 2007. No language restrictions were applied to the searches, but only studies published in English, French or Spanish were reviewed for inclusion.

Study selection

Studies that assessed the accuracy of the following commercial assays were eligible for inclusion: Meridian Premier (enzyme-linked immunosorbent assay (ELISA)); TechLab Tox A/B II (ELISA); TechLab Tox A/B Quik Chek (rapid antigen capture); Remel Xpect (rapid chromatographic immunoassay); bioMérieux VIDAS (EIA); Meridian ImmunoCard (rapid EIA).

Included studies were required to have been conducted in clinical human stool specimens, with a reference standard method for detection of neutralisable C. difficile toxin in cell culture with or without stool culture for C. difficile. The total numbers of patients tested and positive and negative cases had to be reported. Studies were excluded if they were part of the dataset used to define the cut-off for an assay.

Where reported, all studies were conducted in hospital settings in European or North American countries

The authors did not state how many reviewers performed the study selection.

Assessment of study quality

Only studies where the reference standard was applied to all samples were eligible for inclusion (verification biases were avoided). Studies were also assessed for consecutive selection and prospective data collection.

Data extraction

Data to populate 2x2 contingency tables (numbers of true positives, false negatives, false positives and true negatives) were extracted by one reviewer. Study investigators were contacted if further details were needed. Outcome measures reported in the review were sensitivity and specificity, prevalence of positive stools and diagnostic odds ratio.

Methods of synthesis

Median values of sensitivity and specificity with interquartile ranges were reported for each assay. Diagnostic thresholds were set by the manufacturers and did not vary between studies. Between-study heterogeneity in sensitivity and specificity was assessed using the Χ² test. The effect of different reference standards (culture and cell-culture cytotoxicity or cell-culture cytotoxicity alone) on sensitivity and specificity was assessed using the Wilcoxon rank sum test.

Pooled estimates of diagnostic odds ratios, with 95% confidence intervals (CIs), appeared to be calculated and the Kruskal-Wallis test used to compare accuracy between assays (although the authors stated that no formal meta-analysis was conducted). Logistic regression, with only a single factor for commercial assay type, was used to estimate the effect of choice of assay on sensitivity.

Manufacturers' data were excluded.

Results of the review

Eighteen studies reporting 28 datasets were included in the review. The prevalence of positive stool samples ranged from 6% to 53% (median 15%) across these 28 datasets.

There was significant between-study heterogeneity for all assays except TechLab Tox A/B Quik Chek (four studies) and BioMérieux VIDAS (one study).

The median (interquartile range) sensitivities of assays were: TechLab Tox A/B II 0.83 (0.82, 0.85; six studies, n=2,158); Meridian Premier 0.95 (0.86, 0.97; nine studies, n=2,891); TechLab Tox A/B Quik Chek 0.84 (0.81, 0.87; four studies, n=1,307); Remel Xpect 0.82 (0.75, 0.88; two studies, n=520), Meridian ImmunoCard 0.90 (0.87, 0.92; six studies, n=1,982); and BioMérieux VIDAS 0.76 (one study, n=62).

The median (interquartile range) specificities of assays were: TechLab Tox A/B II 0.99 (0.98,1.00; six studies, n=2,158); Meridian Premier 0.97 (0.95, 0.98; nine studies, n=2,891); TechLab Tox A/B Quik Chek 1.00 (0.99, 1.00; four studies, n=1,307); Remel Xpect 0.96 (0.95, 0.98; two studies, n=520); Meridian ImmunoCard 0.99 (0.98, 1.00; six studies, n=1,982); and BioMérieux VIDAS 0.93 (one study, n=62).

The reference method used did not affect sensitivity (p=0.45) or specificity (p=0.36).

There was no evidence for a significant difference in diagnostic odds ratios between any of the assays (p=0.33). Logistic regression suggested significant differences in sensitivity and specificity. Meridian Premier was most likely to give a positive result for true cases and TechLab Quik Chek was most likely to give a negative result for negative samples.

Cost information

The authors estimated that in their hospital a two-stage testing process for C. difficile toxin would result in an additional 500 cell-culture cytotoxicity assays per year at an additional cost of £10,000 to £30,000 per year, but that the cost would be offset by savings in antibiotic costs, use of isolation facilities, reduced infection control burden and reduced cases of C. difficile infection.

Authors' conclusions

Where the prevalence of C. difficile toxin A and B in stool samples was relatively low (<10%), the positive predictive value of the assays was unacceptably low and would vary with assay used and number of samples tested.

CRD commentary

The review presented a clearly stated research question and defined appropriate inclusion criteria. A range of sources were searched to identify relevant studies, but only those published in English, French or Spanish were assessed for inclusion. This left open the possibility of language bias and potential for omission of relevant data. No measures to reduce error and bias in the review process were reported. No assessment of the methodological quality of included studies was made. The extent to which the data presented may have been affected by error in the review process or intrinsic bias could not be assessed. The decision to present ranges of sensitivity and specificity, rather than pooled estimates was appropriate given the observed heterogeneity. However, pooled estimates of diagnostic odds ratios appeared to be used for some analyses. The authors' interpretation of the data, as presented, is reasonable, but should be viewed with caution given the methodological limitations of the review.

Implications of the review for practice and research

Practice: The authors suggested that to improve diagnosis a two-stage testing strategy for C. difficile toxin with an initial highly sensitive rapid screening test to identify positive samples that were then confirmed by a reference method should be used.

Research: The authors made no recommendations for research.

Funding

Not reported.

Bibliographic details

Planche T, Aghaizu A, Holliman R, Riley P, Poloniecki J, Breathnach A, Krishna S. Diagnosis of Clostridium difficile infection by toxin detection kits: a systematic review. Lancet Infectious Diseases 2008; 8(12): 777-784. [PubMed: 18977696]

Original Paper URL

http://www.sciencedirect.com/science/article/pii/S1473309908702330

Indexing Status

Subject indexing assigned by NLM

MeSH

Bacteriological Techniques; Clostridium Infections /diagnosis; Clostridium difficile; Cross Infection /diagnosis; Enterotoxins /analysis; Feces /microbiology; Humans; Polymerase Chain Reaction; Predictive Value of Tests; Reagent Kits, Diagnostic; Sensitivity and Specificity

AccessionNumber

12009101717

Database entry date

23/09/2009

Record Status

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.